VARIABLE-RATE PREDICTIVE RESIDUAL VECTOR QUANTIZER

A major problem with a VQ-based image compression scheme is its codebo ok search complexity, Recently, a predictive residual vector quantizer (PRVQ) was proposed in [4], This scheme has a very low search complex ity, and its performance is very close to that of the predictive vecto r quantizer (PVQ), This letter presents a new VQ scheme called variabl e-rate PRVQ (VR-PRVQ), which is designed by imposing a constraint on t he output entropy of the PRVQ, The proposed VR-PRVQ is found to give a n excellent rate-distortion performance and deafly outperforms the sta te-of-the-art image compression algorithm developed by the Joint Photo graphic Experts Group (JPEG).